Reputation: 3472
I'm trying to implement a neural network in which I need the kernel multiplication with the input only. The dense layer in Tensorflow also adds bias which I am trying to set to zero. From the documentation the only variable that is available to play with is bias_regularizer
. So I tried doing the following:
def make_zero(_):
return np.zeros(21,)
out1 = tf.layers.dense(inputs=codeword, units=21, activation=None, bias_regularizer=make_zero)
But I still see the bias values are not zero. Is there any other method to achieve this?
Upvotes: 5
Views: 4682
Reputation: 59731
See the documentation of tf.layers.dense
, you can use the option use_bias=False
to remove the bias altogether:
out1 = tf.layers.dense(inputs=codeword, units=21, activation=None, use_bias=False)
Upvotes: 8